Characterisation of variable rate video codecs as autoregressive moving average processes for ATM networks

Presents methods of measurements and a procedure for characterising a conditional replenishment (CR) video codec, as described by Chin (see IEEE SAC, SAC-7, p.826, 1989), as an autoregressive moving average process (ARMA). The ARMA process has been adopted because it has been shown by Ramaswami (1988) that the quality of service (cell congestion, cell waiting time in buffer) performance measures of a queueing system are most sensitive to the autocorrelation function of the arrival process. Also the estimated autocorrelation function of the video codec gives a good indication of its functional behaviour. The long term mean, variance and autocovariance function are measured from the cell output of the simulated video codec. These are then used to estimate the parameters of a mathematical model of an ARMA process. The mathematical model consists of a transverse filter of finite order and a recursive filter with a cell interarrival sequence which is assumed to be wide sense stationary (WSS)